Hemogram-based decision tree for predicting the metabolic syndrome and cardiovascular diseases in the elderly.

BACKGROUND This study aimed to build a hemogram-based decision tree to evaluate the association between current probability of metabolic syndrome (MetS) and prediction of future hypertension, type 2 diabetes, and cardiovascular diseases (CVD) risk. METHODS A total of 40,395 elder participants (≥60 years) were enrolled in a standard health examination program in Taiwan from January 1999 to December 2014. A decision tree classification of the presence or absence of MetS at baseline, using age, sex, and hemogram (white blood cell, hemoglobin, and platelet) as independent variables, was conducted for the randomly assigned training (70%) and validation (30%) groups. Participants without MetS at baseline (n = 25,643) were followed up to observe whether they developed MetS, hypertension, type 2 diabetes, or CVD in the future. RESULTS Modest accuracy of the decision tree in the training and validation groups with area under the curves of 0.653 and 0.652, respectively, indicated an acceptable generalizability of results. The predicted probability of baseline MetS was obtained from decision tree analysis. Participants without MetS at baseline were categorized into three equally-sized groups according to the predicted probability. Participants in the third tertile had significantly higher risks of future MetS (hazard ratio 1.40, 95% confidence interval 1.25-1.58); type 2 diabetes (1.46, 1.17-1.83); hypertension (1.14, 1.01-1.28); and CVD (1.21, 1.01-1.44), compared with those in the first tertile. CONCLUSIONS Execution of hemogram-based decision tree analysis can assist in early identification and prompt management of elderly patients at a high risk of future hypertension, type 2 diabetes, and CVD.